Apparatus and Method for Quantification of Pulmonary Function Based on Artificial Intelligence and Medical Images

This application claims priority from Korean Patent Application Nos. 10-2024-0058373 filed on May 2, 2024, and 10-2025-0056647 filed on Apr. 29, 2025, which are incorporated herein by reference in their entireties.

The present disclosure relates to technology for processing, analyzing, and visualizing medical images, and more particularly, to technology for providing a quantitative evaluation to assist the diagnosis of diseases of the lungs using medical images.

The contents described in this section merely provide information about the background art of the present disclosure and do not constitute prior art.

Interstitial lung disease (ILD) is a condition in which changes in the lung parenchyma occur with fibrosis of the lungs. In the past, the disease was diagnosed invasively through lung biopsy, but modern imaging advances have made it possible to diagnose some cases of ILD using imaging alone. However, it is impossible to regenerate damaged lungs to their original state, and the drugs currently on the market do not cure the disease but merely slow down the exacerbation of the disease. Therefore, early diagnosis of the disease is important not only for the quality of life of a patient but also to reduce mortality.

ILD is known to be caused by a combination of environmental factors including smoking and physiological factors including autoimmune diseases. Fortunately, in recent years, there has been an increase in early detection through respiratory screening and chest computed tomography (CT) in comprehensive medical examinations. Accordingly, more patients are actively being examined, and the number of tests is increasing year by year.

To diagnose ILD, a pulmonary function test (PFT) is primarily performed in conjunction with imaging. In performing such a PFT, there are various factors of variation (factors that change data values), and a very large number of instructions are given to control all the factors of variation. Nevertheless, there are many uncontrollable factors, such as patient cooperation and the presence of other medications, which make it difficult to fully trust PFT data. For this reason, image diagnosis and an arterial blood gas test (measurement of oxygen and carbon dioxide tensions) are performed simultaneously with a PFT.

With regard to imaging diagnostics, there has been a technology introduced to automatically diagnose ILD from medical images using artificial intelligence (AI) technology. This is a method of detecting a pattern corresponding to ILD in an image and analyzing the disease from the pattern.

This method is quite useful for detecting ILD but requires a highly sensitive algorithm to detect the disease in its early stages from a medical image. Thus, many disease-like patterns are detected, making it difficult to detect the disease early.

To overcome this limitation, several tests are synthesized. However, it is a complex process for a user to analyze different types of data to draw a comprehensive conclusion, and it is time consuming and costly to train a diagnostician to become proficient. In addition, when limited staff members perform more patient data analysis themselves than in the past, they face the problems of increased fatigue and higher rates of misdiagnosis.

A method of evaluating the progression of a lung disease, such as progressive pulmonary fibrosis, is generally based on the presence of at least two of the following: (1) worsening of the patient's symptoms, (2) evidence of worsening in pulmonary function tests (PFTs), and (3) evidence of worsening in radiological exams.

However, evidence of exacerbation from radiological exams is often subjective, depending on the opinion of the medical worker making the diagnosis. Also, even when texture is automatically extracted from medical images, there are no quantified metrics to ensure that a medical worker's judgment corresponds to an automated analysis result.

Accordingly, example embodiments of the present disclosure are provided to substantially obviate one or more problems due to limitations and disadvantages of the related art.

One object of example embodiments of the present disclosure is to detect/segment clinical findings in a medical image and then estimate a pulmonary function test (PFT) indicator to provide a quantified indicator.

Another object of example embodiments of the present disclosure is to quantify evidence of exacerbation from radiological exams and provide the quantified evidence that is comparable with evidence of exacerbation from a PFT, to reduce differences in diagnosing the progression of a lung disease between individual medical workers.

According to an aspect of the present disclosure, there may be provided a method of quantifying pulmonary function using a medical image, the method comprising: acquiring or receiving a medical image including anatomical information for a lung region of a patient; segmenting at least one abnormal finding region in the lung region of the medical image using an artificial neural network; and predicting a quantification result related to pulmonary function based on a size of the at least one abnormal finding region.

The predicting of the quantification result related to the pulmonary function based on the size of the at least one abnormal finding region may comprise: predicting the quantification result related to the pulmonary function by applying a weight predetermined for type of the at least one abnormal finding region to the size of the at least one abnormal finding.

The method of quantifying pulmonary function using a medical image may further comprise: segmenting the lung region of the medical image into a plurality of anatomical regions, wherein each of the at least one abnormal finding region corresponds to any one of the plurality of anatomical regions.

In this case, the predicting of the quantification result related to the pulmonary function based on the size of the at least one abnormal finding region may comprise: predicting the quantification result related to the pulmonary function by applying weights predetermined for abnormal finding regions each corresponding to the plurality of anatomical regions to sizes of the abnormal finding regions each corresponding to the plurality of anatomical regions.

The predicting of the quantification result related to the pulmonary function based on the size of the at least one abnormal finding region may be performed using a linear regression model learning a function of predicting the quantification result related to the pulmonary function by applying a weight predetermined for type of at least one abnormal finding region to the size of the at least one abnormal finding region.

The predicting of the quantification result related to the pulmonary function based on the size of the at least one abnormal finding region may comprise: generating a prediction value of a pulmonary function test (PFT) result as the quantification result.

The predicting of the quantification result related to the pulmonary function based on the size of the at least one abnormal finding region may comprise: predicting a spirometry result, a diffusing capacity, or a lung volume as the quantification result.

The predicting of the quantification result related to the pulmonary function based on the size of the at least one abnormal finding region may comprise: predicting a forced vital capacity (FVC), a forced expiratory volume in one second (FEV1), or a ratio of FEV1 to FVC (FEV1/FVC) as the spirometry result; and predicting a diffusing capacity of the lung for carbon monoxide (DLCO) as the diffusing capacity.

The predicting of the quantification result related to the pulmonary function based on the size of the at least one abnormal finding region comprises: quantifying an effective lung volume based on the size of the at least one abnormal finding region; and generating the prediction value of the PFT result as the quantification result based on the effective lung volume.

In the method of quantifying pulmonary function using a medical image, the at least one abnormal finding region may include an emphysema region, a consolidation region, a ground-glass opacity (GGO) region, a reticulation region, or a honeycomb region.

The method of quantifying pulmonary function using a medical image may further comprise: generating diagnostic assistance information in regard to interstitial lung disease (ILD) or pneumonia of the lung region based on the quantification result related to the pulmonary function.

The method of quantifying pulmonary function using a medical image may further comprise: visualizing the quantification result.

The predicting of the quantification result related to the pulmonary function based on the size of the at least one abnormal finding region may comprise: acquiring a first weight predetermined for type of at least one abnormal finding region; acquiring a second weight by adjusting the first weight in accordance with whether the patient is treated with an antifibrotic; and predicting the quantification result related to the pulmonary function by applying the second weight to the size of the at least one abnormal finding region.

According to an aspect of the present disclosure, there may be provided an apparatus for quantifying pulmonary function using a medical image, the apparatus comprising: a memory configured to store one or more instructions; and a processor configured to load the one or more instructions from the memory and execute the one or more instructions.

In the apparatus according to the present disclosure, the processor, by executing the one or more instructions, may be configured to: acquire or receive a medical image including anatomical information about a patient's lung region; segment at least one abnormal finding region in the lung region of the medical image using an artificial neural network; and predict a quantification result related to pulmonary function based on size of the at least one abnormal finding region.

The processor may be further configured to predict the quantification result related to the pulmonary function by applying a weight predetermined for type of the at least one abnormal finding region to the size of the at least one abnormal finding region.

The processor may be further configured to: segment the lung region of the medical image into a plurality of anatomical regions; match each of the at least one abnormal finding region to any one of the plurality of anatomical regions, and predict the quantification result related to the pulmonary function based on the sizes of the at least one abnormal finding region by applying weights predetermined for abnormal finding regions each corresponding to the plurality of anatomical regions to sizes of the abnormal finding regions each corresponding to the plurality of anatomical regions.

The processor may be further configured to generate a prediction value of a pulmonary function test (PFT) result as the quantification result.

The processor may be further configured to predict a spirometry result, a diffusing capacity, or a lung volume as the quantification result.

The processor may be further configured to: predict a forced vital capacity (FVC), a forced expiratory volume in one second (FEV1), or a ratio of FEV1 to FVC (FEV1/FVC) as the spirometry result; and predict a diffusing capacity of the lung for carbon monoxide (DLCO) as the diffusing capacity.

In the apparatus of quantifying pulmonary function using a medical image, the at least one abnormal finding region may include an emphysema region, a consolidation region, a ground-glass opacity (GGO) region, a reticulation region, or a honeycomb region.

The processor may be further configured to generate diagnostic assistance information about interstitial lung disease (ILD) or pneumonia of the lung region based on the quantification result related to the pulmonary function.

According to an aspect of the present disclosure, there may be provided a method of quantifying pulmonary function using a medical image, the method comprising: acquiring or receiving a medical image including anatomical information for a lung region of a patient; segmenting at least one abnormal finding region in the lung region of the medical image using an artificial neural network; predicting a first quantification result related to pulmonary function based on a size of the at least one abnormal finding region; and visualizing a second quantification result based on a user input in regard to the at least one abnormal finding region.

The method of quantifying pulmonary function using a medical image may further comprise: generating the second quantification result based on the user input in regard to the at least one abnormal finding region.

The generating the second quantification result may comprise: adjusting the at least one abnormal finding region based on the user input in regard to the at least one abnormal finding region; and generating the second quantification result based on the adjusted at least one abnormal finding region.

The visualizing a second quantification result may comprise: visualizing the first quantification result in addition to a first result of segmentation or classification for the at least one abnormal finding region; and visualizing the second quantification result in addition to a second result of segmentation or classification for the adjusted at least one abnormal finding region.

The user input in regard to the at least one abnormal finding region may include: an input to adjust segmented boundaries of the at least one abnormal finding region, an input to change a classification result for the at least one abnormal finding region, an input to perform a segmentation for the at least one abnormal finding region again, and/or an input to perform a classification for the at least one abnormal finding region again.

According to an aspect of the present disclosure, there may be provided an apparatus for quantifying pulmonary function using a medical image, the apparatus comprising: a memory and a processor, by executing one or more instructions loaded from the memory, may be configured to: acquire or receive a medical image including anatomical information about a patient's lung region; segment at least one abnormal finding region in the lung region of the medical image using an artificial neural network; and predict a first quantification result related to pulmonary function based on a size of the at least one abnormal finding region; and visualize a second quantification result based on a user input in regard to the at least one abnormal finding region.

According to an aspect of the present disclosure, there may be provided a method of quantifying pulmonary function using a medical image, the method comprising: acquiring or receiving a medical image including anatomical information for a lung region of a patient; segmenting at least one abnormal finding region in the lung region of the medical image using an artificial neural network; predicting a quantification result related to pulmonary function based on a user input in regard to the at least one abnormal finding region and further based on a size of the at least one abnormal finding region; and visualizing the quantification result.

According to an aspect of the present disclosure, there may be provided an apparatus for quantifying pulmonary function using a medical image, the apparatus comprising: a memory and a processor, by executing one or more instructions loaded from the memory, may be configured to: acquire or receive a medical image including anatomical information about a patient's lung region; segment at least one abnormal finding region in the lung region of the medical image using an artificial neural network; and predict a quantification result related to pulmonary function based on a user input in regard to the at least one abnormal finding region and further based on a size of the at least one abnormal finding region; and visualize the quantification result.

According to an example embodiment of the present disclosure, it may detect/segment clinical findings including six textures in a medical image and then estimate a PFT indicator to provide a quantified indicator.

According to an example embodiment of the present disclosure, it may quantify evidence of exacerbation from radiological exams and provide the quantified evidence that is comparable with evidence of exacerbation from a PFT.

According to an example embodiment of the present disclosure, it may reduce differences in diagnosing the progression of a lung disease between individual medical workers.

Here, the type of finding region, criteria for detecting/segmenting/classifying/determining the finding region, criteria for determining the size of the finding region, and/or the like may be determined in accordance with a specific disease. In addition, a quantification method for calculating the size of a finding region may be selected among clinically known methods.

Other objects and features of the present disclosure in addition to the above-described objects will be apparent from the following description of embodiments to be given with reference to the accompanying drawings.

The embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings. In the following description, when it is determined that a detailed description of a known component or function may unnecessarily make the gist of the present disclosure obscure, it will be omitted.